DISCLAIMER: This is a very old, rather slow, mostly untested, and completely unmaintained implementation of word2vec for an old course project (i.e., I do not respond to questions/issues). Feel free to fork/clone and modify, but use at your own risk!
A Python implementation of the Continuous Bag of Words (CBOW) and skip-gram neural network architectures, and the hierarchical softmax and negative sampling learning algorithms for efficient learning of word vectors (Mikolov, et al., 2013a, b, c; http://code.google.com/p/word2vec/).
Usage
To train word vectors:
word2vec.py [-h] -train FI -model FO [-cbow CBOW] [-negative NEG]
[-dim DIM] [-alpha ALPHA] [-window WIN]
[-min-count MIN_COUNT] [-processes NUM_PROCESSES]
[-binary BINARY]
required arguments:
-train FI Training file
-model FO Output model file
optional arguments:
-h, --help show this help message and exit
-cbow CBOW 1 for CBOW, 0 for skip-gram
-negative NEG Number of negative examples (>0) for negative sampling,
0 for hierarchical softmax
-dim DIM Dimensionality of word embeddings
-alpha ALPHA Starting learning rate
-window WIN Max window length
-min-count MIN_COUNT Min count for words used to learn <unk>
-processes NUM_PROCESSES Number of processes
-binary BINARY 1 for output model in binary format, 0 otherwise
Each sentence in the training file is expected to be newline separated.
Implementation Details
Written in Python 2.7.6 and NumPy 1.9.1.
Evaluation
Accuracy (%) on the word analogy task compared against the original C implementation (in parentheses). Trained on a preprocessed version of the first 108 bytes of the English Wikipedia dump on March 3, 2006 (http://mattmahoney.net/dc/textdata.html).
| Model | Total | Semantic | Syntactic |
|---|---|---|---|
| CBOW HS | 6.76 (6.90) | 4.86 (3.61) | 7.93 (8.93) |
| CBOW NS | 4.52 (6.72) | 3.94 (3.74) | 4.88 (8.56) |
| Skip-gram HS | 14.76 (14.59) | 11.40 (10.40) | 16.83 (17.18) |
| Skip-gram NS | 8.43 (7.72) | 4.91 (4.62) | 10.62 (9.63) |
References
Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013a). Distributed representations of words and phrases and their compositionality. Advances in Neural Information Processing Systems. http://papers.nips.cc/paper/5021-distributed-representations-of-words-and-phrases-and-their-compositionality.pdf
Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013b). Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. http://arxiv.org/pdf/1301.3781.pdf
Mikolov, T., Yih, W., & Zweig, G. (2013c). Linguistic Regularities in Continuous Space Word Representations. HLT-NAACL. http://msr-waypoint.com/en-us/um/people/gzweig/Pubs/NAACL2013Regularities.pdf